Genetic Configuration Sampling

This is the online support of the project "Genetic Configuration Sampling". This page includes the raw dataset of configuration fault data for three projects(Apache, BusyBox, and Linux) and the prototype of genetic configuration sampling.
If you have interest in our work, please refer to or cite our paper,

Jifeng Xuan, Yongfeng Gu, Zhilei Ren, Xiangyang Jia, and Qingna Fan. 2018. Genetic Configuration Sampling: Learning a Sampling Strategy for Fault Detection of Configurable Systems. In Proceedings of the 5th International Workshop on Genetic Improvement (GI@GECCO 2018), July 15–19, 2018, Kyoto, Japan. ACM, New York, NY, USA, 8 pages.
The usage of the prototype can be found as follows,
java -jar gcs.jar train-test ‹parameter1› ‹parameter2› ...
For example,
java -jar gcs.jar train-test inPathUnion_busybox inPathFault_busybox inPathUnion_linux inPathFault_linux 10 30 90 10 100 20 50 100 10 100
means the input info is as follows,
  • Path (or path union) of training set of configurations: inPathUnion_busybox
  • Path of training set of faults: inPathFault_busybox
  • Path (or path union) of test set of configurations: inPathUnion_linux
  • Path of test set of faults: inPathFault_linux
  • Length of chromosome: 10, Size of the chromosome pool: 30
  • Rate of crossover: 90 %, Rate of mutation: 10 %, Rate of best-chromosome selection: 100 %
  • Maximum number of generation: 20
  • Number of continuously repeating one sampling action (default): 1
  • Number of fitnesses for calculating the average: 50
  • Times of repeating one chromosome (during training) when sampling configurations for a given chromosome: 100
  • Times of repeating one chromosome (during test) when sampling configurations for a given chromosome: 100
  • Times of showing accumulating results from the test set: 10
Note that the path of training (or test) set of configurations supports the path union format. For example, in Linux OS, a path union could be
The dataset is extracted from the original dataset in the following work. Please consider citing the paper for its pioneering and great contribution,

F. Medeiros, C. Kästner, M. Ribeiro, R. Gheyi and S. Apel, "A Comparison of 10 Sampling Algorithms for Configurable Systems," 2016 IEEE/ACM 38th International Conference on Software Engineering (ICSE), Austin, TX, 2016, pp. 643-654. doi: 10.1145/2884781.2884793 URL:
For any question about our work, please contact Jifeng Xuan, Wuhan University.
E-mail: jxuan (at) whu (dot) edu (dot) cn